/*
* Select the desired position for the next block in a file.
*
* we try to mimic what Remy does in inode_getblk/block_getblk
*
* we note: blocknr == 0 means that we're about to allocate either
* a direct block or a pointer block at the first level of indirection
* (In other words, stuff that will go in i_db[] or i_ib[])
*
* blocknr != 0 means that we're allocating a block that is none
* of the above. Then, blocknr tells us the number of the block
* that will hold the pointer
*/
e4fs_daddr_t
ext2_blkpref(struct inode *ip, e2fs_lbn_t lbn, int indx, e2fs_daddr_t *bap,
e2fs_daddr_t blocknr)
{
int tmp;
mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
/* if the next block is actually what we thought it is,
then set the goal to what we thought it should be
*/
if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0)
return ip->i_next_alloc_goal;
/* now check whether we were provided with an array that basically
tells us previous blocks to which we want to stay closeby
*/
if (bap)
for (tmp = indx - 1; tmp >= 0; tmp--)
if (bap[tmp])
return bap[tmp];
/* else let's fall back to the blocknr, or, if there is none,
follow the rule that a block should be allocated near its inode
*/
return blocknr ? blocknr :
(e2fs_daddr_t)(ip->i_block_group *
EXT2_BLOCKS_PER_GROUP(ip->i_e2fs)) +
ip->i_e2fs->e2fs->e2fs_first_dblock;
}
/*
* Allocate a block in the filesystem.
*
* A preference may be optionally specified. If a preference is given
* the following hierarchy is used to allocate a block:
* 1) allocate the requested block.
* 2) allocate a rotationally optimal block in the same cylinder.
* 3) allocate a block in the same cylinder group.
* 4) quadradically rehash into other cylinder groups, until an
* available block is located.
* If no block preference is given the following hierarchy is used
* to allocate a block:
* 1) allocate a block in the cylinder group that contains the
* inode for the file.
* 2) quadradically rehash into other cylinder groups, until an
* available block is located.
*/
int
ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size,
struct ucred *cred, e4fs_daddr_t *bnp)
{
struct m_ext2fs *fs;
struct ext2mount *ump;
int32_t bno;
int cg;
*bnp = 0;
fs = ip->i_e2fs;
ump = ip->i_ump;
mtx_assert(EXT2_MTX(ump), MA_OWNED);
#ifdef INVARIANTS
if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) {
vn_printf(ip->i_devvp, "bsize = %lu, size = %d, fs = %s\n",
(long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt);
panic("ext2_alloc: bad size");
}
if (cred == NOCRED)
panic("ext2_alloc: missing credential");
#endif /* INVARIANTS */
if (size == fs->e2fs_bsize && fs->e2fs->e2fs_fbcount == 0)
goto nospace;
if (cred->cr_uid != 0 &&
fs->e2fs->e2fs_fbcount < fs->e2fs->e2fs_rbcount)
goto nospace;
if (bpref >= fs->e2fs->e2fs_bcount)
bpref = 0;
if (bpref == 0)
cg = ino_to_cg(fs, ip->i_number);
else
cg = dtog(fs, bpref);
bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
ext2_alloccg);
if (bno > 0) {
/* set next_alloc fields as done in block_getblk */
ip->i_next_alloc_block = lbn;
ip->i_next_alloc_goal = bno;
ip->i_blocks += btodb(fs->e2fs_bsize);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bnp = bno;
return (0);
}
nospace:
EXT2_UNLOCK(ump);
ext2_fserr(fs, cred->cr_uid, "filesystem full");
uprintf("\n%s: write failed, filesystem is full\n", fs->e2fs_fsmnt);
return (ENOSPC);
}
/*
* Implement the cylinder overflow algorithm.
*
* The policy implemented by this algorithm is:
* 1) allocate the block in its requested cylinder group.
* 2) quadradically rehash on the cylinder group number.
* 3) brute force search for a free block.
*/
static u_long
ext2_hashalloc(struct inode *ip, int cg, long pref, int size,
daddr_t (*allocator)(struct inode *, int, daddr_t, int))
{
struct m_ext2fs *fs;
ino_t result;
int i, icg = cg;
mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED);
fs = ip->i_e2fs;
/*
* 1: preferred cylinder group
*/
result = (*allocator)(ip, cg, pref, size);
if (result)
return (result);
/*
* 2: quadratic rehash
*/
for (i = 1; i < fs->e2fs_gcount; i *= 2) {
cg += i;
if (cg >= fs->e2fs_gcount)
cg -= fs->e2fs_gcount;
result = (*allocator)(ip, cg, 0, size);
if (result)
return (result);
}
/*
* 3: brute force search
* Note that we start at i == 2, since 0 was checked initially,
* and 1 is always checked in the quadratic rehash.
*/
cg = (icg + 2) % fs->e2fs_gcount;
for (i = 2; i < fs->e2fs_gcount; i++) {
result = (*allocator)(ip, cg, 0, size);
if (result)
return (result);
cg++;
if (cg == fs->e2fs_gcount)
cg = 0;
}
return (0);
}
/*
* Common code for mount and mountroot.
*/
static int
ext2_mountfs(struct vnode *devvp, struct mount *mp)
{
struct ext2mount *ump;
struct buf *bp;
struct m_ext2fs *fs;
struct ext2fs *es;
struct cdev *dev = devvp->v_rdev;
struct g_consumer *cp;
struct bufobj *bo;
struct csum *sump;
int error;
int ronly;
int i, size;
int32_t *lp;
ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
/* XXX: use VOP_ACESS to check FS perms */
DROP_GIANT();
g_topology_lock();
error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
g_topology_unlock();
PICKUP_GIANT();
VOP_UNLOCK(devvp, 0);
if (error)
return (error);
/* XXX: should we check for some sectorsize or 512 instead? */
if (((SBSIZE % cp->provider->sectorsize) != 0) ||
(SBSIZE < cp->provider->sectorsize)) {
DROP_GIANT();
g_topology_lock();
g_vfs_close(cp);
g_topology_unlock();
PICKUP_GIANT();
return (EINVAL);
}
bo = &devvp->v_bufobj;
bo->bo_private = cp;
bo->bo_ops = g_vfs_bufops;
if (devvp->v_rdev->si_iosize_max != 0)
mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
if (mp->mnt_iosize_max > MAXPHYS)
mp->mnt_iosize_max = MAXPHYS;
bp = NULL;
ump = NULL;
if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
goto out;
es = (struct ext2fs *)bp->b_data;
if (ext2_check_sb_compat(es, dev, ronly) != 0) {
error = EINVAL; /* XXX needs translation */
goto out;
}
if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
(es->e2fs_state & E2FS_ERRORS)) {
if (ronly || (mp->mnt_flag & MNT_FORCE)) {
printf(
"WARNING: Filesystem was not properly dismounted\n");
} else {
printf(
"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
error = EPERM;
goto out;
}
}
ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
/*
* I don't know whether this is the right strategy. Note that
* we dynamically allocate both an ext2_sb_info and an ext2_super_block
* while Linux keeps the super block in a locked buffer.
*/
ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
M_EXT2MNT, M_WAITOK);
ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
M_EXT2MNT, M_WAITOK);
mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
goto out;
/*
* Calculate the maximum contiguous blocks and size of cluster summary
* array. In FFS this is done by newfs; however, the superblock
* in ext2fs doesn't have these variables, so we can calculate
* them here.
*/
ump->um_e2fs->e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
if (ump->um_e2fs->e2fs_maxcontig > 0)
ump->um_e2fs->e2fs_contigsumsize =
MIN(ump->um_e2fs->e2fs_maxcontig, EXT2_MAXCONTIG);
else
ump->um_e2fs->e2fs_contigsumsize = 0;
if (ump->um_e2fs->e2fs_contigsumsize > 0) {
size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
lp = ump->um_e2fs->e2fs_maxcluster;
sump = ump->um_e2fs->e2fs_clustersum;
for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
*lp++ = ump->um_e2fs->e2fs_contigsumsize;
sump->cs_init = 0;
sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
}
}
brelse(bp);
bp = NULL;
fs = ump->um_e2fs;
fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
/*
* If the fs is not mounted read-only, make sure the super block is
* always written back on a sync().
*/
fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
if (ronly == 0) {
fs->e2fs_fmod = 1; /* mark it modified */
fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */
}
mp->mnt_data = ump;
mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
MNT_ILOCK(mp);
mp->mnt_flag |= MNT_LOCAL;
MNT_IUNLOCK(mp);
ump->um_mountp = mp;
ump->um_dev = dev;
ump->um_devvp = devvp;
ump->um_bo = &devvp->v_bufobj;
ump->um_cp = cp;
/*
* Setting those two parameters allowed us to use
* ufs_bmap w/o changse!
*/
ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
if (ronly == 0)
ext2_sbupdate(ump, MNT_WAIT);
/*
* Initialize filesystem stat information in mount struct.
*/
MNT_ILOCK(mp);
mp->mnt_kern_flag |= MNTK_MPSAFE | MNTK_LOOKUP_SHARED |
MNTK_EXTENDED_SHARED;
MNT_IUNLOCK(mp);
return (0);
out:
if (bp)
brelse(bp);
if (cp != NULL) {
DROP_GIANT();
g_topology_lock();
g_vfs_close(cp);
g_topology_unlock();
PICKUP_GIANT();
}
if (ump) {
mtx_destroy(EXT2_MTX(ump));
free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
free(ump->um_e2fs->e2fs, M_EXT2MNT);
free(ump->um_e2fs, M_EXT2MNT);
free(ump, M_EXT2MNT);
mp->mnt_data = NULL;
}
return (error);
}
/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*
*/
static u_long
ext2_dirpref(struct inode *pip)
{
struct m_ext2fs *fs;
int cg, prefcg, cgsize;
u_int avgifree, avgbfree, avgndir, curdirsize;
u_int minifree, minbfree, maxndir;
u_int mincg, minndir;
u_int dirsize, maxcontigdirs;
mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED);
fs = pip->i_e2fs;
avgifree = fs->e2fs->e2fs_ficount / fs->e2fs_gcount;
avgbfree = fs->e2fs->e2fs_fbcount / fs->e2fs_gcount;
avgndir = fs->e2fs_total_dir / fs->e2fs_gcount;
/*
* Force allocation in another cg if creating a first level dir.
*/
ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref");
if (ITOV(pip)->v_vflag & VV_ROOT) {
prefcg = arc4random() % fs->e2fs_gcount;
mincg = prefcg;
minndir = fs->e2fs_ipg;
for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
mincg = cg;
minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
}
for (cg = 0; cg < prefcg; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
mincg = cg;
minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
}
return (mincg);
}
/*
* Count various limits which used for
* optimal allocation of a directory inode.
*/
maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg);
minifree = avgifree - avgifree / 4;
if (minifree < 1)
minifree = 1;
minbfree = avgbfree - avgbfree / 4;
if (minbfree < 1)
minbfree = 1;
cgsize = fs->e2fs_fsize * fs->e2fs_fpg;
dirsize = AVGDIRSIZE;
curdirsize = avgndir ? (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0;
if (dirsize < curdirsize)
dirsize = curdirsize;
maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255);
maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR);
if (maxcontigdirs == 0)
maxcontigdirs = 1;
/*
* Limit number of dirs in one cg and reserve space for
* regular files, but only if we have no deficit in
* inodes or space.
*/
prefcg = ino_to_cg(fs, pip->i_number);
for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
return (cg);
}
for (cg = 0; cg < prefcg; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
return (cg);
}
/*
* This is a backstop when we have deficit in space.
*/
for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
return (cg);
for (cg = 0; cg < prefcg; cg++)
if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
break;
return (cg);
}
